Safety Events in Children’s Hospitals During the COVID-19 Pandemic Free

We conducted a retrospective cohort study of pediatric discharges (inpatient or observation, children aged 0–18 years) in the Pediatric Health Information System (PHIS) database. Preliminary PHIS analyses revealed a significant decline in volume of children’s hospital encounters beginning on March 15, 2020. We therefore defined our study periods as March 15 to May 31, 2020 (during COVID-19) and March 15 to May 31, 2017–2019 (pre-COVID-19). The PHIS database contains administrative data, including 41 International Classification of Diseases 10th Revision diagnosis and procedure codes on discharges from 52 tertiary care children’s hospitals across the United States.⁸  The PHIS database represents ∼20% of pediatric hospitalizations to children’s hospitals nationally. Data quality is maintained jointly between the Children’s Hospital Association (Lenexa, KS) and participating hospitals. The 45 hospitals that provided data throughout the study period were included. Sample size was determined by these selection criteria.

The primary outcome was the rate of each PDI, calculated by using the AHRQ PDIs v2020 International Classification of Diseases 10th Revision, Clinical Modification and Procedure Coding System software (numerator and denominator definitions and predictor and confounder selection are detailed by the AHRQ).⁹  We calculated the overall and specific PDI rates within the pre-COVID-19 and COVID-19 periods, which were reported as PDI events per 1000 discharges. The overall PDI rate was calculated by pooling the individual measures’ numerators and denominators (ensuring that discharges were not double counted). We assessed patient characteristics (age, sex, and race and/or ethnicity), encounter type (medical or surgical, based on the assigned All Patient Refined Diagnosis Related Groups), payer, and disposition. We assessed for medical complexity using the number of complex chronic conditions (CCCs).¹⁰  We calculated the case-mix index for each hospitalization using the Hospitalization Resource Intensity Score for Kids (H-RISK), an All Patient Refined Diagnosis Related Groups–based measure of relative resource intensity specifically for pediatric hospital care and a proxy measure for clinical severity (a higher score indicates higher clinical severity).^11–13  Given rigorous PHIS data validation efforts, none of our included predictor variables had >1% missing, except race, which was missing in 4.2% of the patients. All patients missing race data were grouped into the “other” race category.

Categorical demographic and clinical characteristics were summarized by using frequencies and percentages and compared between periods by using χ² tests for categorical variables and Student’s t tests for the H-RISK. Because the PDIs all have distinct population definitions, we tested them as 8 independent hypotheses rather than combined. Risk-adjusted PDI rates were compared between periods by using generalized linear mixed-effects models with a binary distribution and random intercepts for each hospital. For nonsurgical PDIs, models were adjusted for age, sex, race and/or ethnicity, encounter type, payer, CCC subtype, and H-RISK. For surgical PDIs, to account for differences related to surgical procedures, rather than adjusting for H-RISK, we replicated established methods for surgical safety risk adjustment from the National Surgical Quality Improvement Program by developing a linear risk score of an event (ie, the linear predictor from a logistic regression) for each principal procedure grouped using the AHRQ’s clinical classifications software (see Table 1 for example).^14–16  The linear risk scores were developed independently for each surgical PDI by using the Healthcare Cost and Utilization Project’s 2016 Kids’ Inpatient Database and applied to the PHIS at the discharge level. All statistical analyses were performed by using SAS version 9.4 (SAS Institute, Inc, Cary, NC). P values <.05 were considered statistically significant. This study was determined to be nonhuman subjects research by the institutional review board.

There were 399 113 pediatric hospital discharges in the pre-COVID-19 period (March to May 2017–2019) and 88 140 in the COVID-19 period (March 15 to May 31, 2020). During COVID-19, there was a lower discharge volume across all hospitals, with a median difference of −35.8% (interquartile range: −28.0% to −39.9%). During COVID-19, the majority of hospital discharges involved patients who were 0 to 4 years old (52% vs 49% pre-COVID-19), were female (51% vs 49% pre-COVID-19), and had government insurance (51% vs 53% pre-COVID-19). During COVID-19, the majority had a medical encounter (76% vs 77% pre-COVID-19), was discharged from the hospital (93% vs 93% pre-COVID-19), and had no CCCs (54% vs 55% pre-COVID-19). Compared with pre-COVID-19, hospital discharges during COVID-19 more often involved patients with surgical encounters, patients who were in the younger or older age categories, patients who were female, patients who had a nongovernment payer type, and patients with CCCs (Table 2; all P values significant at <.001). The mean H-RISKs were 2.96 (SE 5.9) and 3.47 (SE 6.9) in the pre-COVID-19 and COVID-19 periods, respectively.

In unadjusted analyses, the overall PDI rate per 1000 discharges during COVID-19 was higher than the pre-COVID-19 PDI rate (6.39 vs 5.05 [P < .001]; Table 3). For specific PDIs, during COVID-19, there were higher rates of postoperative sepsis (13.12 vs 8.41; P < .001) and central venous catheter–related bloodstream infections (2.0 vs 1.58; P = .024) compared with the pre-COVID-19 period. After adjustment, there was no difference seen in the overall PDI rate, although there was significantly increased odds of postoperative sepsis (adjusted odds ratio 1.28; 95% confidence interval [CI] 1.04–1.56) during COVID-19 compared with the pre-COVID-19 period. We examined the most frequent procedures driving the increase in postoperative sepsis during COVID-19, and disproportionately higher rates were observed for “other heart procedures” (eg, insertion of implantable heart assist device; excess events = 4 per 1000); organ transplant (excess events = 3 per 1000); “other diagnostic cardiovascular” procedures (eg, insertion of implantable heart assist system into heart; excess events = 3 per 1000); incision of pleura, thoracentesis, and chest drainage (excess events = 3 per 1000); and heart valve procedures (excess events = 2 per 1000) (Table 1).

In this retrospective cohort study of children’s hospital discharges, we found an increased risk for overall safety events (ie, PDIs), postoperative sepsis, and central venous catheter bloodstream infections during the COVID-19 pandemic. However, after adjusting for patient and clinical factors, we found that the only PDI with increased odds during COVID-19 was postoperative sepsis. In light of the ongoing pandemic and morbidity and mortality associated with postoperative sepsis,¹⁷  efforts are needed to further understand this alarming finding to mitigate risk for children requiring surgery during the COVID-19 pandemic.

Because many elective procedures were canceled during the early COVID-19 pandemic, it is possible that patients undergoing surgeries were more critically ill compared with the pre-COVID-19 population. However, after risk adjusting for clinical severity, the increase for postoperative sepsis remained significant. Several factors may have contributed to this finding, including delays in presentation, diagnosis, and management. In multicenter risk adjustment studies, researchers have identified several risk factors for postoperative sepsis, including wound class (eg, dirty wounds), need for oxygen support, and higher American Society of Anesthesiologists classification.¹⁴  These factors are associated with more severe disease at time of operation, which could reflect delayed presentation as well as delayed diagnosis and definitive management. More severe presentations associated with delay in management has also been associated with intussusception, volvulus, bowel obstruction, and infections.^18–20  Studies in children with appendicitis reveal that lack of health insurance and decreased health care use can lead to delayed diagnosis and increased risk for complications, such as postoperative sepsis.^21–23 

To our knowledge, there is no literature on trends in hospital safety events during the COVID-19 pandemic, although there are anecdotal reports of increased hospital-acquired infections.²⁴  After hospital presentation, there are many in-hospital factors impacted by the COVID-19 pandemic that could lead to increased risk for postoperative sepsis. These include delay in surgical management due to preoperative COVID-19 testing, modified workflow leading to delays in diagnostic testing, need for increased sterilization, and infection control precautions impacting operating room flow. It is also conceivable that changes in workflow and personnel and need for personal protective equipment for hospital staff could lead to delays in postoperative management. This could lead to delayed recognition and treatment of hospital-acquired infections that could then progress to sepsis. Although we were unable to ascertain the relative influence of pre- and in-hospital factors on the increase in postoperative sepsis, previous data have revealed that the COVID-19 pandemic has impacted health care access, including decreased health care use, thus leading to potential increased risk for postoperative sepsis.²⁵ 

Limitations of the study include the use of administrative data: the data set lacked certain clinical and other factors that may have impacted safety events in the pediatric hospital during COVID-19. We attempted to minimize this limitation using hierarchical modeling to adjust for PDI clustering among hospitals and used previously established methods for surgical PDI risk adjustment. Our risk adjustment methods were also based on administrative data and may not fully account for true clinical severity. Our findings are limited to freestanding children’s hospitals and thus may not be generalizable to other settings. Because our study was conducted in the early months of the COVID-19 pandemic, further study is needed to understand the full extent of changes in PDIs during the COVID-19 pandemic.

After adjusting for patient and clinical factors, we found increased odds for postoperative sepsis among children hospitalized during COVID-19. Further study is warranted to better understand the underlying drivers for this trend, especially regarding pre- and in-hospital factors that may influence the risk of sepsis after surgical procedures. This information can then inform efforts to improve timely presentation, diagnosis, and management in children requiring surgery to reduce morbidity and mortality in this population. Given the concern that the COVID-19 pandemic will persist for months to years, as well as concern for risk of future pandemics, continued efforts are warranted to ensure the safety of hospitalized children.